It seems to be incoming link anchor text which kicks the process off, if you have unnatural anchor text with lots of money keywords from web2.0s, social bookmarks, forum profiles then google takes a closer look at the site.
If you then have thin content or affiliate outlinks then the site can take a nosedive.
For example I have two UK sites in the camping niche.
Both sites are built exactly the same with fairly decent onsite content 10-20 pages, not big but big enough with Amazon outlinks.
Each site caters to a size - one site sells the small size and one site sells the larger size.
I threw some web2.0 backlinks and social bookamrks at the site that sells the small product but never got around to doing the same thing with the large product on the other site.
And of course the site selling the small product has been zapped, the larger product on the other site is untouched.
They are totally the same in all other respects both built on wordpress using Reviewazon to generate affiliate commissions.
I have other examples of this, in a nutshell I think it is the incoming anchor text which starts the spam evaluation.